Upload BAD classifier - Layer 13 - Acc: 69.83%

README.md

@@ -0,0 +1,60 @@
+---
+license: apache-2.0
+tags:
+- fairsteer
+- bias-detection
+- debiasing
+- tinyllama
+library_name: pytorch
+---
+
+# BAD Classifier for FairSteer - TinyLlama-1.1B
+
+This is a Biased Activation Detection (BAD) classifier trained for the FairSteer framework.
+
+## Model Details
+
+- **Base Model**: TinyLlama/TinyLlama-1.1B-Chat-v1.0
+- **Task**: Binary classification (Biased vs Unbiased activations)
+- **Training Data**: BBQ dataset with balanced sampling
+- **Best Layer**: 13
+- **Validation Accuracy**: 69.83%
+- **Architecture**: Simple linear classifier (FairSteer-aligned)
+
+## Usage
+```python
+import torch
+import json
+
+# Load model
+model = torch.load("pytorch_model.bin")
+with open("config.json", "r") as f:
+    config = json.load(f)
+
+# Use for bias detection
+# Input: activation vector from LLM layer 13
+# Output: probability of being unbiased
+```
+
+## Training Details
+
+- **Samples**: 24,276 balanced samples
+- **Class Distribution**: 50% BIASED, 50% UNBIASED
+- **Training Method**: FairSteer-aligned labeling
+- **Training Date**: 2025-11-16
+
+## Citation
+
+If you use this model, please cite the FairSteer paper:
+```bibtex
+@article{fairsteer,
+  title={FairSteer: Inference-Time Debiasing for Large Language Models},
+  author={[Authors]},
+  journal={[Journal]},
+  year={2024}
+}
+```
+
+## License
+
+Apache 2.0

config.json

@@ -0,0 +1,11 @@
+{
+  "input_dim": 2048,
+  "layer_idx": 13,
+  "base_model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
+  "best_val_accuracy": 0.6983113673805601,
+  "training_method": "balanced_sampling",
+  "mmlu_used": false,
+  "balanced": true,
+  "samples_per_class": 12138,
+  "training_date": "2025-11-16T04:04:08.618908"
+}

layer_comparison.json

@@ -0,0 +1,10 @@
+{
+  "layer_results": {
+    "12": 0.6620675453047776,
+    "13": 0.6983113673805601,
+    "14": 0.675658978583196,
+    "15": 0.6668039538714992
+  },
+  "best_layer": 13,
+  "best_accuracy": 0.6983113673805601
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f8a84d404f75bc46126eae477b527a7fab0a01861742f8a28b6082e775249f7
+size 9788

training_history.json

@@ -0,0 +1,442 @@
+{
+  "train_loss": [
+    0.6879778053807181,
+    0.6775284423641516,
+    0.6690367475228747,
+    0.6618725060434469,
+    0.6559289151313499,
+    0.6506586861168442,
+    0.646369316563179,
+    0.6424463251592202,
+    0.6392711363786526,
+    0.6364438485902811,
+    0.6339698331608954,
+    0.6315967033623913,
+    0.6295402436128729,
+    0.6277056113576054,
+    0.6258861495954983,
+    0.6244743900613362,
+    0.6230971551702639,
+    0.62169047979567,
+    0.6204631547235436,
+    0.6193978288146667,
+    0.6182792164515761,
+    0.6171552758015516,
+    0.6161465937764711,
+    0.6152513820282831,
+    0.6144809927311788,
+    0.6135997878289738,
+    0.6126996867811397,
+    0.6118564106163354,
+    0.6112918919809814,
+    0.6104648998787916,
+    0.6097356335510309,
+    0.6090582758955557,
+    0.6085019414751464,
+    0.6077826469492102,
+    0.607138771094942,
+    0.6066668764201791,
+    0.6061470598442793,
+    0.605538748209312,
+    0.6050183299027314,
+    0.6044908564687389,
+    0.60413318992766,
+    0.6035735612549079,
+    0.6032753308573172,
+    0.6030734250400633,
+    0.6029054640495937,
+    0.602635905327684,
+    0.6023594082936445,
+    0.6022516398670499,
+    0.6019733499547603,
+    0.6017306041152789,
+    0.6015415748901642,
+    0.6013156224844016,
+    0.6011463924981538,
+    0.6010031350854742,
+    0.6009114068519443,
+    0.6007802489243379,
+    0.600656607185407,
+    0.6006341064749736,
+    0.6005215973613437,
+    0.6004067136870599,
+    0.6002956700128581,
+    0.600224838382061,
+    0.6001235851544184,
+    0.6000645204872108,
+    0.6000029330887092,
+    0.5999573764668434,
+    0.5999164908725373,
+    0.5998703609779861,
+    0.5998168483453235,
+    0.5997992949063451,
+    0.5997742460144782,
+    0.599759944054911,
+    0.5997198893407591,
+    0.5996954524185582,
+    0.5996735734502515,
+    0.5996641645593574,
+    0.5996489726920083,
+    0.5996351324075528,
+    0.5996226743615619,
+    0.5996095734325178,
+    0.5996015670984585,
+    0.599592800660924,
+    0.5995885842737774,
+    0.5995822438741933,
+    0.5995784508212096,
+    0.599573129605805
+  ],
+  "train_acc": [
+    0.5770339855818744,
+    0.6386199794026777,
+    0.6523686920700309,
+    0.6524201853759012,
+    0.6525746652935118,
+    0.6551493305870237,
+    0.6554067971163748,
+    0.6582904222451081,
+    0.6573120494335737,
+    0.6575695159629248,
+    0.6573120494335737,
+    0.658908341915551,
+    0.659989701338826,
+    0.6606076210092688,
+    0.6621524201853759,
+    0.6619979402677652,
+    0.662821833161689,
+    0.6633882595262616,
+    0.666735324407827,
+    0.6659629248197735,
+    0.6672502574665293,
+    0.6675077239958805,
+    0.6686405767250257,
+    0.669773429454171,
+    0.6692070030895984,
+    0.6705458290422245,
+    0.6714727085478888,
+    0.6720906282183317,
+    0.6719361483007209,
+    0.6730690010298661,
+    0.6737384140061792,
+    0.6751802265705459,
+    0.6751802265705459,
+    0.6752832131822863,
+    0.6754891864057673,
+    0.6759011328527291,
+    0.6768795056642637,
+    0.67826982492276,
+    0.6780123583934089,
+    0.6783213182286303,
+    0.6776519052523172,
+    0.6781668383110195,
+    0.678475798146241,
+    0.6790937178166838,
+    0.6791452111225541,
+    0.6792996910401647,
+    0.6790937178166838,
+    0.6801235839340886,
+    0.6805355303810504,
+    0.679557157569516,
+    0.6811534500514933,
+    0.6791967044284243,
+    0.6800205973223481,
+    0.6799176107106076,
+    0.680638516992791,
+    0.6796086508753862,
+    0.6825952626158599,
+    0.6796086508753862,
+    0.6805355303810504,
+    0.6812049433573636,
+    0.6798661174047373,
+    0.681513903192585,
+    0.6805870236869207,
+    0.680638516992791,
+    0.6812049433573636,
+    0.6814624098867147,
+    0.6814109165808445,
+    0.6800720906282184,
+    0.6816683831101956,
+    0.6807415036045315,
+    0.6814109165808445,
+    0.6812564366632338,
+    0.6813594232749742,
+    0.681307929969104,
+    0.6813594232749742,
+    0.6818228630278064,
+    0.681513903192585,
+    0.6815653964984552,
+    0.6816168898043254,
+    0.6815653964984552,
+    0.6814624098867147,
+    0.6814624098867147,
+    0.6814624098867147,
+    0.6814624098867147,
+    0.681513903192585,
+    0.6814624098867147
+  ],
+  "val_loss": [
+    0.6814653777212052,
+    0.6711087451735281,
+    0.6624498628509496,
+    0.654977040883933,
+    0.6487816975772676,
+    0.643314992478102,
+    0.6388996045122036,
+    0.6344917111184687,
+    0.6310531487182102,
+    0.6277697638192719,
+    0.6250191223876677,
+    0.6225961511571089,
+    0.6205937972375156,
+    0.6181881586841539,
+    0.616502619261011,
+    0.6145167099388663,
+    0.6129057519321975,
+    0.6114355510306319,
+    0.6103526705375809,
+    0.6088014954986447,
+    0.6076663166922832,
+    0.6065283853689568,
+    0.6055230404046261,
+    0.6044670277608089,
+    0.6035148835260746,
+    0.6026557307856086,
+    0.6018306628679718,
+    0.6008309255518277,
+    0.6000772093037601,
+    0.5992732013666257,
+    0.5985813128103731,
+    0.597856211112398,
+    0.5972650096286856,
+    0.5970398709055227,
+    0.596298783084706,
+    0.5954595113900387,
+    0.594806127630702,
+    0.5943971040614158,
+    0.5937728629473603,
+    0.5935884918373145,
+    0.5927945107368897,
+    0.5924836442340933,
+    0.5923015638277598,
+    0.5921205562068367,
+    0.5918064597607444,
+    0.5915167164763273,
+    0.5912831992254619,
+    0.5910642328725811,
+    0.5908447867366784,
+    0.5906345579141055,
+    0.5906986622676818,
+    0.5902115965792925,
+    0.5901051024235846,
+    0.5900135116875662,
+    0.589906981003147,
+    0.5898232718276035,
+    0.5898531813794345,
+    0.5896186457434439,
+    0.5895111150364507,
+    0.5894343130278233,
+    0.5893136986594224,
+    0.5892664673300905,
+    0.5891904199535803,
+    0.5891255477114996,
+    0.5890939631414963,
+    0.5890515091195334,
+    0.5890037779753055,
+    0.5889491310229608,
+    0.5889198634730533,
+    0.5889005954615171,
+    0.5888786426878839,
+    0.5888424443059541,
+    0.5888212450842881,
+    0.5887979702069692,
+    0.5887906491461853,
+    0.5887780335825397,
+    0.588766589097961,
+    0.5887586958522263,
+    0.5887460385555098,
+    0.5887359151926622,
+    0.5887309737024826,
+    0.5887248832272148,
+    0.5887190115314339,
+    0.5887122939799527,
+    0.5887037247566651,
+    0.5886992343764329
+  ],
+  "val_acc": [
+    0.6441515650741351,
+    0.6696869851729819,
+    0.6577429983525536,
+    0.6690691927512356,
+    0.6649505766062603,
+    0.6742174629324547,
+    0.6762767710049423,
+    0.675658978583196,
+    0.6709225700164745,
+    0.6760708401976936,
+    0.6754530477759473,
+    0.6801894563426688,
+    0.6742174629324547,
+    0.6816309719934102,
+    0.6771004942339374,
+    0.6828665568369028,
+    0.6845140032948929,
+    0.6838962108731467,
+    0.6878088962108732,
+    0.6847199341021417,
+    0.6838962108731467,
+    0.6861614497528831,
+    0.6884266886326195,
+    0.6865733113673805,
+    0.6869851729818781,
+    0.6908978583196046,
+    0.6861614497528831,
+    0.6890444810543658,
+    0.6890444810543658,
+    0.6904859967051071,
+    0.693163097199341,
+    0.6919275123558485,
+    0.689662273476112,
+    0.685337726523888,
+    0.6970757825370676,
+    0.6906919275123559,
+    0.6919275123558485,
+    0.689662273476112,
+    0.6925453047775947,
+    0.6884266886326195,
+    0.6941927512355849,
+    0.6954283360790774,
+    0.6937808896210873,
+    0.692339373970346,
+    0.6960461285008237,
+    0.6974876441515651,
+    0.6960461285008237,
+    0.69666392092257,
+    0.6970757825370676,
+    0.6972817133443163,
+    0.6925453047775947,
+    0.6970757825370676,
+    0.6968698517298187,
+    0.6970757825370676,
+    0.6968698517298187,
+    0.6983113673805601,
+    0.6956342668863262,
+    0.6970757825370676,
+    0.6968698517298187,
+    0.69666392092257,
+    0.6970757825370676,
+    0.6956342668863262,
+    0.6964579901153213,
+    0.6970757825370676,
+    0.6968698517298187,
+    0.6968698517298187,
+    0.6970757825370676,
+    0.69666392092257,
+    0.69666392092257,
+    0.6968698517298187,
+    0.6970757825370676,
+    0.6974876441515651,
+    0.6970757825370676,
+    0.6972817133443163,
+    0.6968698517298187,
+    0.6968698517298187,
+    0.6968698517298187,
+    0.69666392092257,
+    0.69666392092257,
+    0.6964579901153213,
+    0.6964579901153213,
+    0.6964579901153213,
+    0.6964579901153213,
+    0.6964579901153213,
+    0.6968698517298187,
+    0.6964579901153213
+  ],
+  "lr": [
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    3.125e-06,
+    3.125e-06,
+    3.125e-06,
+    3.125e-06,
+    3.125e-06,
+    3.125e-06,
+    1.5625e-06,
+    1.5625e-06,
+    1.5625e-06,
+    1.5625e-06,
+    1.5625e-06,
+    1.5625e-06
+  ]
+}